Regional Resource Adequacy Studies

Appendix F: Resource Adequacy F 2017 PSE Integrated Resource Plan Regional Resource Adequacy Studies The results and data from these three studies of regional load/resource balance were used in the preparation of the 2017 PSE IRP. Contents 1. NORTHWEST POWER AND CONSERVATION COUNCIL (NPCC) Pacific Northwest Power Supply Adequacy Assessment for 2021 Published September 27, 2016 (attached) 2. PACIFIC NORTHWEST UTILITIES CONFERENCE COMMITTEE () Northwest Regional Forecast of Power Loads and Resources 2017-2026 Published April 2016 (attached) 3. BONNEVILLE POWER ADMINISTRATION (BPA) 2016 Pacific Northwest Loads and Resources Study Published December 22, 2016 Access this document at the following links: Summary https://www.bpa.gov/power/pgp/whitebook/2016/2016-wbk-loads-and- Resources-Summary-20161222.pdf Energy Analysis https://www.bpa.gov/power/pgp/whitebook/2016/2016-wbk-technical- Appendix-Volume-1-Energy-Analysis-20161222.pdf Capacity Analysis https://www.bpa.gov/power/pgp/whitebook/2016/2016-wbk-technical- Appendix-Volume%20-2-Capacity-Analysis-20161222.pdf F - 1 PSE 2017 IRP

CONTENTS Forward... 3 Executive Summary... 4 The Council s Resource Adequacy Standard... 6 Recent Adequacy Assessments... 7 2021 Resource Adequacy Assessment... 8 Sensitivity Analysis...10 Monthly Analysis...12 Curtailment Statistics...13 Other Adequacy Metrics...20 Assumptions...22 Future Assessments...26 Table 1: History of Adequacy Assessment... 8 Table 2: Load and SW Market Impacts to LOLP (121 MW new DR)...10 Table 3: Load and SW Market Uncertainty LOLP Map Existing (500 MW new DR)...11 Table 4: Load and SW Market Uncertainty LOLP Map Existing (1,257 MW new DR)...11 Table 5: Sensitivity Loss of Gas Supply/Market Friction (Loss of 650 MW IPP Resource)...11 Figure 1: LOLP by Month...12 Table 6: Expected Resource Dispatch for 2021...13 Table 7: 2021 Simulated Curtailment Statistics...14 Figure 2: Curtailment Event Duration Probability...15 Figure 3: Event Duration Frequency (1-hour block incremental)...16 Figure 4: Event Duration Frequency (2-hour block incremental)...16 Figure 5: Event Duration Frequency (various time blocks)...17 Figure 6: Annual Unserved Energy Probability...18 Figure 7a: Worst-Hour Unserved Energy Probability...19 Figure 7b: Worst-Hour Unserved Energy Probability (Blow Up)...19 Table 8: Adequacy Metric Definitions...21 Table 9: Annual Adequacy Metrics (Base Case)...21 Table 10: Monthly Adequacy Metrics (Base Case)...22 Table 11: Assumptions used for the 2021 Adequacy Assessment...23 Table 12: Standby Resource Assumptions Peak (MW)...23 Table 13: Standby Resource Assumptions Energy (MW-hours)...24 Table 14: Within-hour Balancing Reserves Incremental (MW)...25 Table 15: Within-hour Balancing Reserves Decremental (MW)...26 2

FORWARD This document summarizes the Northwest Power and Conservation Council s assessment of the adequacy of the power supply for the 2021 operating year (October through September). In 2011, the Council adopted the annual loss-of-load probability (LOLP) as the measure for power supply adequacy and set the maximum value at 5 percent. For a power supply to be deemed adequate, the likelihood (LOLP) of a shortfall (not necessarily an outage) occurring anytime in the year being examined cannot exceed 5 percent. Other adequacy metrics that measure the size of potential shortages, how often they occur and how long they last, also provide valuable information to planners as they consider resource expansion strategies. This report provides that information along with other statistical data derived from Council analyses. The Council, with the help of the Resource Adequacy Advisory Committee, produced the data in the charts and tables. The format and content of this report continue to be under development. We would like to know how useful this report is for you. For example, is the format appropriate? Would you like to see different types of output? Please send your comments, suggestions and questions to John Fazio at (jfazio@nwcouncil.org). The Council is improving its adequacy model (GENESYS), in particular the hourly hydroelectric system dispatch simulation, and expects to complete the work by 2018. In addition, the Council has initiated a process to review its current adequacy standard. Staff and RAAC members have been asked to review the viability of the current metric (LOLP) and threshold (5 percent). This review should consider similar efforts going on in other parts of the United States, namely through the IEEE Loss-of-Load-Expectation Working Group and the North American Electric Reliability Corporation (NERC). Cover photo courtesy of SOAR Oregon. 3

EXECUTIVE SUMMARY The Pacific Northwest s power supply should be adequate through 2020. However, with the planned retirements of four Northwest coal plants 1 by July of 2022, the system will no longer meet the Council s adequacy standard and will have to acquire nearly 1,400 megawatts of new capacity in order to maintain that standard. This result assumes that the region will meet the Council s energy efficiency targets, as identified in the Seventh Power Plan. Thus, it is imperative that we continue to implement cost-effective energy efficiency programs. Beyond energy efficiency, Northwest utilities have been steadily working to develop replacement resource strategies and have reported about 550 megawatts of planned generating capacity by 2021. 2 These strategies will include the next most cost-effective and implementable resources, which may include additional energy efficiency, demand response or new generating resources. The Council will reassess the adequacy of the power supply next year to monitor the region s progress in maintaining resource adequacy. In 2011, the Northwest Power and Conservation Council adopted a regional adequacy standard to provide an early warning should resource development fail to keep pace with demand growth. The standard deems the power supply to be inadequate if the likelihood of a power supply shortfall (referred to as the loss-of-load probability or LOLP) is higher than 5 percent. The LOLP for the region s power supply should stay under the 5 percent limit through 2020. In 2021, with the loss of 1,330 megawatts of capacity from the Boardman and Centralia 1 coal plants (slated to retire in December of 2020), the LOLP rises to 10 percent. 3 In this scenario, the region will need a little over 1,000 megawatts of new capacity to maintain adequacy. Should the Colstrip 1 and 2 coal plants (307 megawatts committed to serve regional demand) also retire before 2021, 4 the LOLP grows to just over 13 percent and the region s adequacy need grows to about 1,400 megawatts of new capacity. These results are based on a stochastic analysis that simulates the operation of the power supply over thousands of different combinations of river flow, wind generation, forced outages, and temperatures. Since last year s assessment for 2021, which resulted in an 8 percent LOLP, 1 Centralia 1 (670 megawatts) and Boardman (522 megawatts) are scheduled to retire by December 2020, Colstrip 1 and 2 (154 megawatts each) are to be retired no later than July of 2022 and Centralia 2 (670 megawatts) is expected to retire by 2025. 2 From the Pacific Northwest Utility Conference Committee s 2016 Northwest Regional Forecast (NRF). 3 Boardman and Centralia 1 coal plants are scheduled to retire in December 2020. However, because the Council s operating year runs from October 2020 through September 2021, these two plants would be available for use during the first three months of the 2021 operating year. For this scenario, the LOLP is 7.6 percent. The Council must take into account the long-term effects of these retirements, and therefore uses the more generic study that has both plants out for the entire operating year. 4 Currently there is no indication that Colstrip plants 3 and 4 will be retired earlier than expected. 4

the region s load forecast has slightly decreased 5 and no new resources have been added. This year s LOLP assessment for 2021 has grown to 10 percent because it included all regional balancing reserve requirements instead of only the federal system reserves assumed in last year s analysis. The conclusions made above assume that future demand will stay on the Council s medium load forecast path and that only a fixed amount of imported generation from the Southwest is available. If demand growth were to increase rapidly and if the availability of imports were to drop, the LOLP could grow as high as 30 percent and the region s adequacy needs could grow to 2,600 megawatts or more. But these extreme cases are not very likely to occur. Resource acquisition plans to bring the 2021 power supply into compliance with the Council s standard will vary depending on the types of new generating resources or demand reduction programs that are considered. In all likelihood, utilities will use some combination of new generation and load reduction programs to bridge the gap. This analysis does not provide a strategy to maintain an adequate, efficient, economical, and reliable power supply. The Council s Seventh Power Plan outlines a resource strategy to ensure an adequate power supply for 2021. Northwest utilities, as reported in the Pacific Northwest Utilities Conference Committee s 2016 Northwest Regional Forecast, show about 550 megawatts of planned generating capacity for 2021. However, these planned resources are not sited and licensed and are therefore not included in the 2021 adequacy assessment. As conditions change over the next few years, we expect utilities to revise their resource acquisition strategies to invest in new resources, which include energy efficiency and demand response. 5 This year s assessment included a hybrid load forecasting method that is different from past forecasts. This was done to insure that the load forecast used for the adequacy assessment was consistent with the one used for the development of the Council s Seventh Power Plan. The RAAC will evaluate this new load forecast in detail prior to next year s assessment for 2022. 5

THE COUNCIL S RESOURCE ADEQUACY STANDARD In 2011, the Northwest Power and Conservation Council adopted a regional adequacy standard to provide an early warning should resource development fail to keep pace with demand growth. The standard deems the power supply to be inadequate if the likelihood of a power supply shortfall five years in the future is higher than 5 percent. The Council assesses adequacy using a stochastic analysis to compute the likelihood of a supply shortfall. It uses a chronological hourly simulation of the region s power supply over many different future combinations of stream flows, temperatures, wind generation patterns and forced generator outages. We only count existing generating resources, and those expected to be operational in the study year, along with targeted energy efficiency savings. The simulation also assumes a fixed amount of market resource availability, both from inside and outside of the region. The power supply is deemed to be adequate if the likelihood of a shortfall (referred to as the loss of load probability or LOLP) is less than or equal to 5 percent. If the supply is deemed inadequate, the Council estimates how much additional capacity and energy generating capability is required to bring the system s LOLP back down to 5 percent. However, the standard is not intended to provide a resource-planning target because it assesses only one of the Council s criteria for developing a power plan. The Council s mandate is to develop a resource strategy that provides an adequate, efficient, economic and reliable power supply. There is no guarantee that a power supply that satisfies the adequacy standard will also be the most economical or efficient. Thus, the adequacy standard should be thought of as simply an early warning to test for sufficient resource development. Because the computer model used to assess adequacy (GENESYS) cannot possibly take into account all contingency actions that utilities have at their disposal to avert an actual loss of service, a non-zero LOLP should not be interpreted to mean that real curtailments will occur. Rather, it means that the likelihood of utilities having to take extraordinary and costly measures to provide continuous service exceeds the tolerance for such events. Some emergency utility actions are captured in the LOLP assessment through a post-processing program that simulates the use of what the Council has termed standby resources. Standby resources are demand-side actions and small generators that are not explicitly modeled in the adequacy analysis. They are mainly composed of demand response measures, load curtailment agreements and small thermal resources. Demand response measures are typically expected to be used to help lower peak-hour demand during extreme conditions (e.g. high summer or low winter temperatures). These resources only have a capacity component and provide only a very limited amount of energy (i.e. they cannot be dispatched for more than a few hours at a time). The effects of demand response measures that have already been implemented are assumed to be reflected in the Council s load forecast. 6

New demand response measures that have no operating history and are therefore not accounted for in the load forecast are classified as part of the set of standby resources. Load curtailment actions, which are contractually available to utilities to help reduce peak hour load, and small generating resources may also provide some energy assistance. However, they are not intended to be used often and are, therefore not modeled explicitly in the simulations. The energy and capacity capabilities of these non-modeled resources are aggregated along with the demand response measures mentioned above to define the total capability of standby resources. A post-processing program uses these capabilities to adjust the simulated curtailment record and calculate the final LOLP. RECENT ADEQUACY ASSESSMENTS Table 1 below illustrates the evolving nature of the effort to better quantify power supply adequacy. Since 1998, when the Council began using stochastic methods to assess adequacy, the power supply and, to some extent the methodology, have changed significantly, sometimes making it difficult to compare annual assessments. And, while this evolution is likely to continue, the Council believes that the current standard and methodology will be sufficiently stable to create a history of adequacy evaluations that can be used to record trends over time. The Council recognizes that the power system of today is very different from that of 1980, when the Council was created by Congress. In particular, the ever increasing generation from variable energy resources, such as solar and wind, have added a greater band of uncertainty with regard to providing an adequate supply. This has led to a greater need in the ability to model hourly operations, especially for the hydroelectric system. Toward this end, the Council is currently in the process of redeveloping its adequacy model (GENESYS) to add more precision to the simulation of hydroelectric generation. The thrust of this effort is to improve the hourly operation simulation by adding a better representation of unit commitment, balancing reserve allocation and moving to a plant-specific hourly hydroelectric simulation (the current model simulates hourly hydroelectric generation in aggregate for the region). These enhancements, expected to be completed by 2018, could likely change the results in a significant way. It will require an extensive vetting effort to ensure that the results of the redeveloped model are a better representation of real-life operations. It will be important to identify the effects of the model enhancements to the resulting adequacy assessments and separate them from the effects of real load and resource changes. 7

Table 1: History of Adequacy Assessment Year Analyzed Operating Year LOLP Observations 2010 2015 5% Was part of the Council s 6 th Power Plan 2012 2017 7% 2014 2019 6% 2015 2020 5% Imports decreased from 3,200 to 1,700 MW, load growth 150 amw per year, only 114 MW of new thermal capacity Load growth 120 amw per year, over 600 MW new generating capacity, increased imports by 800 MW Lower load forecast, 350 amw of additional EE savings 2015 2021 8% 2016 2021 10% 2016 2021 13% Early estimate (BPA INC/DEC only) Loss of Boardman and Centralia 1 (~1,330 MW) 2021 loads lower than last year s forecast regional INC/DEC reduces hydro peaking Same as above but with Colstrip coal plants 1 and 2 retired (307 MW assigned to serve the region) 2021 RESOURCE ADEQUACY ASSESSMENT The Pacific Northwest s power supply is expected to be adequate through 2020. However, with the planned retirements of four Northwest coal plants by July of 2022, the system will no longer meet the Council s adequacy standard (LOLP at 13 percent) and will have to acquire nearly 1,400 megawatts of new capacity in order to reduce the LOLP to the 5 percent standard. This result assumes that the Council s energy efficiency targets, as identified in the Seventh Power Plan, will be achieved. In 2021, with the loss of 1,330 megawatts of capacity from the Boardman and Centralia 1 coal plants (slated to retire in December of 2020), the LOLP rises to 10 percent. 6 In this scenario, the region will need a little over 1,000 megawatts of new capacity to maintain adequacy. Should the Colstrip 1 and 2 coal plants (307 megawatts committed to serve regional demand) also retire 6 Boardman and Centralia 1 coal plants are scheduled to retire in December 2020. However, because the Council s operating year runs from October 2020 through September 2021, these two plants would be available for use during the first three months of the 2021 operating year. For this scenario, the LOLP is 7.6 percent. The Council must take into account the long-term effects of these retirements, and therefore uses the more generic study that has both plants out for the entire operating year. 8

before 2021, the LOLP grows to just over 13 percent and the region s adequacy need grows to about 1,400 megawatts of new capacity. The conclusions made above assume that future demand will stay on the Council s medium load forecast path and that only a fixed amount of imported generation from the Southwest is available. If demand growth were to increase rapidly and if the availability of imports were to drop, the LOLP could grow as high as 26 percent and the region s adequacy needs could grow to 2,600 megawatts or more. But this extreme case is not very likely to occur. Two future uncertainties not modeled explicitly in GENESYS are long-term (economic) load growth and variability of the out-of-region market supply. Long-term load growth is bounded by the Council s high and low load forecasts, which cover roughly 85 percent of the expected load range. Variation in SW market supply is influenced by future resource development in California and by the ability to transfer surplus energy into the Northwest. By 2021, California is scheduled to retire 2,641 megawatts of its coastal water-cooled thermal power plants, and nearly 10,000 megawatts will either be retired or replaced over the next 10 years. In addition, in 2012 California lost 2,200 megawatts of San Onofre Nuclear Generating Station capacity. 7 However, according to an Energy GPS report, California surplus is expected to greatly exceed the south-to-north intertie transfer capability during Northwest winter peakload hours. Based on a look at historical monthly south-to-north transfer availability (BPA data), it appears that the maximum transfer capability hovers around 4,500 megawatts with a 95 percent chance of being at least 3,400 megawatts. The Council chose to set the maximum transfer capability from California into the Northwest to the 3,400 megawatt value. In spite of the results of the Energy GPS survey of available California surplus, and supported by the Resource Adequacy Advisory Committee, the Council chose to limit California import availability to no more than 2,500 megawatts during peak hours in the winter and to 3,000 megawatts during off-peak hours year round. The on-peak imports are defined as a spot market resource, which can be acquired during the hour of need. The off-peak imports are defined as a purchase ahead resource, which can be acquired during the light-loads hours prior to an anticipated peak-hour shortfall. To investigate the potential impacts of different combinations of economic load growth and California import availability, scenario analyses were performed. In one extreme case, with high load growth and no California import, the loss of load probability would be 26 percent. Fortunately, this scenario is not very likely. At the other end of extreme cases, with low load growth and maximum winter import availability, the loss of load probability drops to about 2 percent. Table 2 illustrates how LOLP changes as both long-term load growth and SW imports vary. 7 By 2025 the Diablo Canyon nuclear plant (2,200 megawatts) is expected to close. 9

Table 2: Load and SW Market Impacts to LOLP (121 MW new DR) Import 3400 MW 2500 MW 1700 MW High Load 22.1 24.2 26.2 Med Load 7.8 9.9 12.0 Low Load 1.9 3.7 5.6 Sensitivity Analysis Sensitivity analyses are useful in helping to understand how results may change as particular input assumptions vary. We have already seen, in the section above, how LOLP changes as economic load growth and SW market assumptions vary. In this section, the sensitivity of LOLP to additional demand response and to a loss of gas supply is investigated. Tables 3 and 4 show how LOLP changes as more demand response is added to the power supply. 8 Studies run to produce the results in these tables are identical to those run to produce the results in Table 2, with the exception that more demand response was added to each. In Table 3, an additional 379 megawatts of demand response was added to all the studies (for a total of 500 megawatts of new demand response). In Table 4 an additional 1,136 megawatts (or a total of 1,257 megawatts) of new demand response was added. As evident in the results summarized in these tables, demand response can be a very effective resource toward maintaining an adequate supply. Studies using the Council s Regional Portfolio Model, during the development of the Seventh Power Plan, indicated that up to about 1,300 megawatts of new demand response resource could be cost effective relative to other options to maintain adequacy. Unfortunately, the infrastructure and experience needed to acquire that much new demand response is not as well developed as for energy efficiency programs, thus there remains uncertainty whether this level of new demand response would actually be implementable by 2021. The Council has encouraged utilities to continue to investigate and develop means to more easily acquire cost-effective demand response resources both for winter and summer needs. 8 It should be emphasized that demand response is exclusively a capacity provider with very limited energy contributions. As such, it may not be the best solution to offset longer-term curtailments (e.g. those that last over the 16 peak load hours of the day). 10

Table 3: Load and SW Market Uncertainty LOLP Map Existing (500 MW new DR) Import 3400 MW 2500 MW 1700 MW High Load 15.9 18.5 20.4 Med Load 5.5 7.7 9.5 Low Load 1.4 3.0 5.0 Table 4: Load and SW Market Uncertainty LOLP Map Existing (1,257 MW new DR) Import 3400 MW 2500 MW 1700 MW High Load 7.6 10.0 12.5 Med Load 2.6 4.7 6.7 Low Load 0.4 1.9 3.5 Table 5: Sensitivity Loss of Gas Supply/Market Friction (Loss of 650 MW IPP Resource) Import Base Case IPP Loss + 121 MW DR IPP Loss + 500 MW DR IPP Loss + 1257 MW DR High Load 24.2 30.0 23.1 13.3 Med Load 9.9 13.2 9.6 6.1 Low Load 3.7 5.4 4.5 2.9 Table 5 summarizes the sensitivity of LOLP to a loss of Northwest market supply due to a shortage of fuel (gas). The Northwest has about 3,000 megawatts (nameplate) of independent power producer (IPP) generating capability. Council adequacy assessments assume that all of that capability is available for Northwest use during winter months but only 1,000 megawatts is available during summer months (due to competition with SW utilities). These sensitivity studies examined how much the LOLP increases due to a loss of 650 megawatts of IPP generation during winter and about a 220 megawatt loss of IPP generation during summer. As is evident in that table, a loss of Northwest market has a similar effect on LOLP (making it bigger) as does the loss of SW market supply. This type of analysis could also be thought of as a surrogate for a market friction sensitivity analysis. Market friction is commonly thought of as a decrease in market access due to transmission limitations or due to more conservative operations by utilities during periods of short supply (e.g. utilities may hold more generating capability in reserve during certain conditions) or a combination of both. This type of analysis will be important to investigate further for future adequacy assessments. 11

Monthly Analysis Currently, the Council s adequacy standard sets a 5 percent maximum threshold for annual loss of load probability. This standard has been very useful in the past, especially compared to older deterministic methods, to aid the region in maintaining an adequate power supply. However, with the addition of more and more variable energy generation resources, such as wind and solar, and with the anticipated large increase in solar rooftop development, an annual metric may no longer be the best measure for adequacy. Figure 1 below shows the monthly LOLP values for both the reference case and the case with Colstrip 1 and 2 also retired. It is clear from this figure that the region has both winter and summer adequacy issues. For the reference case, the highest monthly LOLP values still appear mostly in winter but when the two Colstrip plants are also removed, the late summer LOLP value exceeds the winter month values. It is important to differentiate by month (or at least by season) in order to find optimum resource acquisition strategies. For example, some demand response programs are only available in winter or in summer. It should be noted that the sum of monthly LOLP values will not equal the annual value because the annual value counts simulations with at least one curtailment event regardless of when it occurs. A simulation with multiple events, say one in January and one in August, would count the same for the annual LOLP value as a simulation with only a January event or only an August event. Monthly values for other adequacy metrics are summarized in that section of this report. Figure 1: LOLP by Month 4.0 3.5 3.0 Base Case Without Colstrip 2.5 2.0 1.5 1.0 0.5 0.0 Oct Nov Dec Jan Feb Mar Apr 1 Apr 2 May Jun Jul Aug 1 Aug 2 Sep LOLP (%) 12

Table 6 summarizes the average monthly dispatch for groups of resources, namely wind, coal, gas, nuclear and SW market. This table shows the monthly dispatch for the reference case and for the case with the Colstrip 1 and 2 coal plant retirement and the difference. With the added loss of Colstrip 1 and 2, as expected, gas generation and SW market purchases go up to cover, as best they can, the loss of the coal generating capability. Obviously, the shift in the dispatch for these resources is not sufficient to offset the loss of the Colstrip plants as evident in the increase in curtailment events and the increase in the LOLP. Table 6: Expected Resource Dispatch for 20219 2021 Base Case OCT NOV DEC JAN FEB MAR AP1 AP2 MAY JUN JUL AU1 AU2 SEP Wind 1203 1248 1201 1312 1296 1560 1767 1862 1751 1704 1571 1454 1342 1150 Coal 3254 2754 2861 2225 1828 1484 1557 801 467 670 1784 2862 3259 3533 Gas 2710 1184 1310 1356 1043 752 776 563 494 560 847 1596 2048 2439 Nuclear 1034 1039 1070 1075 1128 1076 1071 1066 1076 1053 1077 1067 1110 1055 SW Market 487 505 603 593 343 174 211 55 9 24 88 249 338 403 2021 No Colstrip OCT NOV DEC JAN FEB MAR AP1 AP2 MAY JUN JUL AU1 AU2 SEP Wind 1203 1248 1201 1312 1296 1560 1767 1862 1751 1704 1571 1454 1342 1150 Coal 3027 2561 2672 2054 1718 1410 1474 777 466 649 1679 2700 2986 3224 Gas 2895 1271 1409 1425 1093 785 819 574 495 571 898 1711 2197 2625 Nuclear 1034 1039 1070 1075 1128 1076 1071 1066 1076 1053 1077 1067 1110 1055 SW Market 524 569 674 648 383 202 240 64 10 28 99 277 375 440 No Colstrip - Base OCT NOV DEC JAN FEB MAR AP1 AP2 MAY JUN JUL AU1 AU2 SEP Wind 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Coal -227-193 -189-171 -110-74 -83-24 -1-21 -105-162 -273-309 Gas 185 87 99 69 50 33 43 11 1 11 51 115 149 186 Nuclear 0 0 0 0 0 0 0 0 0 0 0 0 0 0 SW Market 37 64 71 55 40 28 29 9 1 4 11 28 37 37 Curtailment Statistics 9 These studies for the 2021 operating year included no maintenance for the region s sole nuclear plant, which is in error. The 2-year maintenance schedule for the Columbia Generating Station has that plant out of service for about a 2 month period during odd years. So, these studies should have shown zero capability for nuclear during May and June. Since no curtailments are expected during these months, even with the shutdown of the nuclear plant, the resulting LOLP values would remain unchanged. 13

Sometimes, simply looking at simulation results can provide insight into the behavior of the power system. Table 7 below summarizes a few statistics for the curtailment events reported in our analysis. All adequacy studies were run with 6,160 simulations. Besides looking at curtailment statistics, it may also be of great use to examine what conditions existed during the time of each shortfall. Thus, a record of all curtailment events along with the values for the four random variables used in the analysis will be provided in a separate spreadsheet (available on the Council s website). The four random variables displayed in the spreadsheet are; Water supply, as a percentage of monthly runoff volume Temperature, as a percentage of that day s historical temperature range Wind generation, based on historical wind capacity factors from BPA s wind fleet Forced outage conditions Some attempts have been made to correlate shortfall events with the occurrence of certain temperatures, water conditions, wind generation patterns and forced outages, but unfortunately without much success. This is an area of study that is being explored further and may produce better results once the GENESYS model has been enhanced to model plant-specific hourly hydroelectric operations. Table 7: 2021 Simulated Curtailment Statistics Statistic Units Number of simulations 6,160 Number Simulations with a curtailment 610 Number Loss of load probability (LOLP) 10 Percent Number of curtailment events 2,374 Number Number of events per year 0.4 Events/year Average event duration 11 Hours Average event magnitude 12,700 MW-hours Average event peak curtailment 1,200 MW Expected curtailed hours per year (LOLH) 2.4 Hours Expected un-served energy (EUE) 2,500 MW-hours Events with duration of 1 to 2 hours 11 Percent Duration of 1 to 4 hours 20 Percent Duration of 1 to 6 hours 28 Percent Duration of 1 to 12 hours 49 Percent Duration of 1 to 14 hours 56 Percent Duration of 1 to 16 hours 86 Percent Duration greater than 16 hours 14 Percent Highest likely duration (15 to 16 hours) 30 Percent 14

Figure 2 can be used to examine the likelihood for particular duration curtailment events. In that figure, the y-axis represents the duration for an event and the x-axis represents the probability of an event with that duration (or greater) of occurring. For example, in Figure 2 the 50 th percentile duration (median value) is about 13 hours. 10 This means that we expect a 50 percent chance of observing a curtailment event of 13 hours or more. 25 Figure 2: Curtailment Event Duration Probability 20 Hours 15 10 5 0 0% 20% 40% 60% 80% 100% Probability over all Events Figure 3 shows the same information in a different way. In that figure, the y-axis represents the percent of times that an event of particular duration occurs in the study. This is commonly referred to as a frequency distribution chart. For example, the most likely duration for an event is 16 hours. From Figure 3 a 16-hour duration event has about a 25 percent chance of occurring. The second most likely duration for an event is 18 hours. This result is not surprising since GENESYS will attempt to uniform any shortfall it sees across all the high-load hours of the day. Figure 4 shows the same information but the curtailment durations have been combined into 2- hour bins (as opposed to single hour bins in Figure 3). Figure 4 simply highlights the result that most event durations are between 15 and 18 hours. And, finally, Figure 5 provides more of a cumulative probability for event duration. 10 Note that the median duration is 13 hours while the average duration is 11 hours. This is because the distribution of event durations is not symmetric. 15

Figure 3: Event Duration Frequency (1-hour block incremental) 30% 25% 20% Percent 15% 10% 5% 0% 1 hour 2 hours 3 hours 4 hours 5 hours 6 hours 7 hours 8 hours 9 hours 10 hours 11 hours 12 hours 13 hours 14 hours 15 hours 16 hours 17 hours 18 hours 19 hours 20 hours Duration of Curtailment (hours) Figure 4: Event Duration Frequency (2-hour block incremental) 35% 30% 25% Percent 20% 15% 10% 5% 0% 1-2 3-4 5-6 7-8 9-10 11-12 13-14 15-16 17-18 19-20 Duration of Curtailment (hours) 16

Figure 5: Event Duration Frequency (various time blocks) 60% 50% 40% 30% 20% 10% 0% 1 to 2 hours 1 to 4 hours 1 to 6 hours 1 to 12 hours Over 12 hours The point at which these curves cross the horizontal axis would represent the LOLP except that these data were plotted prior to the implementation of standby resources. 11 By applying the effects of standby resources to the reference case results, the LOLP drops from a little over 13 percent down to the final value of 9.9 percent. In other words, if we could modify the curtailment record for that case to shows the effects of standby resources, the resulting probability curve would shift down and cross the horizontal axis at 9.9 percent. Doing the same for the Colstrip retirement case drops the LOLP to a little over 13 percent. Figure 6 displays the annual unserved energy probability over all games for both the reference case and the Colstrip retirement case. The total unserved energy for each of the 6,160 games is summed up and then sorted from highest to lowest. Those results are then graphed in Figure 6. The vertical axis represents the amount of annual unserved energy and the horizontal axis represents the likelihood of observing a particular amount of annual unserved energy or more. From Figure 6, without the effects of standby resources, it appears that there is about a 13 percent 12 chance of observing a game with at least one curtailment (this is where the curve in Figure 6 crosses the horizontal axis). The probability curve for the Colstrip retirement case crosses the horizontal axis at about 17.7 percent. 11 This is a simplification of the actual process, which takes into account monthly results. 12 Remember this result is prior to adding the effects of standby resources. 17

Figure 6: Annual Unserved Energy Probability 500000 Base No Colstrip Megawatt-Hours 400000 300000 200000 100000 13.3 % 17.7 % 0 0% 5% 10% 15% 20% Probability of Exceeding Figure 7a displays the worst-hour unserved energy probability for all games for both the reference case and the Colstrip retirement case. This figure is similar to Figure 6 but plots the worst (highest) single-hour unserved energy for each game, instead of the annual unserved energy. As expected, the probability curves in this figure cross the horizontal axis at the same percentage values as the curves in the annual unserved energy chart (Figure 6). The curves in this figure can be used to estimate the amount of additional capacity needed to make the power supply adequate (not including the effects of standby resources). By looking at a blown-up section of Figure 7a, shown in Figure 7b, it becomes easier to see how much new capacity is required to shift the entire curve down so that it crosses the horizontal axis at the 5 percent Council adequacy limit. For the reference case, it requires a little over 1,800 megawatts of new capacity (simply draw a straight line up from the 5 percent point on the horizontal axis to the curve and then draw a straight line to the left to see where it would cross the vertical axis). Recall that these data have not been adjusted for standby resources, which contribute a little over 600 megawatts of capacity in winter. Thus, the estimate for required new capacity in addition to the standby resource contribution to maintain adequacy is about 1,200 megawatts. For the Colstrip retirement case, the needed amount of new capacity is about 1,500 megawatts. These values, however, are only estimates because they lump the curtailment events from all months together. Results from the more accurate analytical approach (which also include the effects of standby resources) show a need of about 1,040 megawatts and 1,400 megawatts of new capacity to maintain adequacy for the reference case and Colstrip retirement case, respectively. 18

It should be noted that it requires both new capacity and energy additions to move the 2021 LOLP down to the Council s 5 percent standard. Analysis indicates that the greatest need for the 2021 supply is addition of capacity, however, simply adding capacity with no energy will not result in an adequate supply. Each new resource has at least some energy providing capability, some more than others. For example, demand response programs can provide a lot of capacity but cannot be dispatched for long periods of time and therefore, provide only a very limited amount of energy. Wind resources, on the other hand, can provide a great deal of energy but can only be counted on to provide about 5 percent of their nameplate capacity toward peaking needs. This is why the Council uses its Regional Portfolio Model, which knows the energy and capacity contributions of all new resources, to develop a resource strategy that will lead to an adequate supply. Figure 7a: Worst-Hour Unserved Energy Probability 7000 6000 Base No Colstrip Megawatt-Hours 5000 4000 3000 2000 1000 0 0% 5% 10% 15% 20% Probability of Exceeding Figure 7b: Worst-Hour Unserved Energy Probability (Blow Up) 19

Megawatt-Hours 2200 2150 2100 2050 2000 1950 1900 1850 1800 1750 1700 Base No Colstrip 4% 5% 6% Probability of Exceeding Other Adequacy Metrics Other adequacy metrics help planners better understand the magnitude, frequency and duration of curtailments. These other metrics provide valuable information to planners as they consider resource expansion strategies. Table 8 below defines some of the more commonly used probabilistic metrics used to examine power supply adequacy and Table 9 provides the regional assessments of these metrics for 2017, 2019, 2020 and 2021. While the Council has been using an annual LOLP metric to assess adequacy for nearly a decade, it became evident during the development of the Seventh Power Plan that monthly (or at least quarterly) values are essential to ensure a truly adequate supply. This is because resources can provide different energy and capacity contributions over each quarter. Also, the characteristics of potential shortfalls can vary by season. Thus, the Council s Regional Portfolio Model required quarterly adequacy reserve margins to develop more cost effective resource expansion strategies. The calculation of quarterly adequacy reserve margins requires quarterly adequacy targets. Recognizing this, the Council added an action item to reevaluate and amend its existing adequacy standard. Table 10 provides monthly values for LOLP and other adequacy metrics. The North American Electric Reliability Corporation (NERC) instigated an adequacy assessment pilot program in 2012. It asked that each sub-region in the United States provide three adequacy measures; 1) expected loss of load hours, 2) expected unserved energy and 3) normalized expected unserved energy (EUE divided by load). This effort is a good first step toward standardizing how adequacy is assessed across the United States but it falls far short of 20

establishing adequacy thresholds for these metrics. It may, in fact, be impossible to set thresholds because power supplies can vary so drastically across regions. Table 8: Adequacy Metric Definitions Metric LOLP (%) CVaR Energy (MW-hours) CVaR Peak (MW) EUE (MW-hours) LOLH (Hours) PGC (%) Description Loss of load probability = number of games with a problem divided by the total number of games Conditional value at risk, energy = average annual curtailment for 5% worst games Conditional value at risk, peak = average single-hour curtailment for worst 5% of games Expected unserved energy = total curtailment divided by the total number of games Loss of load hours = total number of hours of curtailment divided by total number of games Percent of games with curtailment prior to implementing standby resources Table 9: Annual Adequacy Metrics (Base Case) Metric 2017 2019 2020 2021 Units LOLP 6.6 5.9 4.7 9.9 Percent CVaR - Energy 99,000 59,200 50,589 46,378 MW-hours CVaR - Peak 4,000 3,337 2,949 2,185 MW EUE 5,000 3,000 2,536 2,482 MW-hours LOLH 2.7 1.7 1.5 2.4 Hours/year PGC 9.7 8.3 6.4 13.6 Percent 21

Table 10: Monthly Adequacy Metrics (Base Case) Month LOLP Peak % LOLP Energy % Overall LOLP % EUE MW-Hours LOLH Hours Annual 9.9 1.8 9.9 2,482 2.4 Oct 1.7 0.3 1.7 240 0.5 Nov 0.7 0.1 0.7 170 0.1 Dec 2.5 0.5 2.5 768 0.6 Jan 2.2 0.6 2.2 930 0.6 Feb 0.3 0.2 0.3 105 0.1 Jul 0 0 0 1 0 Au1 1.4 0.2 1.4 102 0.2 Au2 1.9 0.4 2 146 0.3 Sep 0.5 0.1 0.6 21 0.1 Assumptions The methodology used to assess the adequacy of the Northwest power supply assumes a certain amount of reliance on non-utility supplies within the region and imports from California. The Northwest electricity market includes independent power producer (IPP) resources. The full capability of these resources, 2,943 megawatts, is assumed to be available for Northwest use during winter months. However, during summer months, due to competition with California utilities, the Northwest market availability is limited to 1,000 megawatts. Other assumptions used for the 2021 adequacy assessment are shown in Table 11 through Table 15. Table 11 summarizes assumptions for load, energy efficiency savings and out-ofregion market availability. Tables 12 and 13 provide the energy and capacity contributions for standby resources. Tables 14 and 15 provide the monthly incremental and decremental balancing reserves that were assumed. To the extent possible, the hydroelectric system was used to carry these reserves. Using the Council s hourly hydroelectric optimization program (TRAP model), a portion of the peaking capability and minimum generation at specific hydroelectric projects was reserved to support the within-hour balancing needs. Unfortunately, not all balancing reserves could be assigned to the hydroelectric system. The remaining reserves should be assigned to other resources but the current adequacy model does not have that capability. This is one of the major enhancements targeted in the GENESYS redevelopment process. 22

Table 11: Assumptions used for the 2021 Adequacy Assessment Item Quarter 4 Quarter 1 Quarter 2 Quarter 3 Mean Load (amw) 21,234 20,975 18,813 19,987 Peak Load (MW) 33,768 33,848 26,504 28,302 DSI Load (amw) 338 338 338 338 Mean EE (amw) 1,545 1,574 1,274 1,208 Peak EE (MW) 2,660 2,660 1,680 1,680 Spot Imports (MW) 2,500 2,500 0 0 Purchase Ahead (MW) 3,000 3,000 3,000 3,000 Table 12: Standby Resource Assumptions Peak (MW) Item Quarter 4 Quarter 1 Quarter 2 Quarter 3 Exist DR 220 220 781 781 Exist Emergency Gen 266 266 266 266 Total Existing 486 486 1047 1047 Planned DR 121 121 0 0 Total Exist + Planned 607 607 1047 1047 Min DR (from the RPM) 379 379 468 468 13 Total Exist + Plan + Min 986 986 1515 1515 Expected DR (from RPM) 1,136 1,136 1,178 1,178 Total Exist + Plan + Expect 1,743 1,743 2,225 2,225 13 These are existing summer demand response programs. 23

Table 13: Standby Resource Assumptions Energy (MW-hours) Item Quarter 4 Quarter 1 Quarter 2 Quarter 3 Exist DR 37,250 37,250 69,542 69,542 Exist Emergency Gen 5,800 5,800 5,800 5,800 Total Existing 43,050 43,050 75,342 75,342 Planned DR 6,050 6,050 0 0 Total Exist + Planned 49,100 49,100 75,342 75,342 Min DR (from the RPM) 18,950 18,950 23,400 23,400 Total Exist + Plan + Min 68,050 68,050 98,742 98,742 Expected DR (from RPM) 56,800 56,800 58,900 58,900 Total Exist + Plan + Expect 105,900 105,900 134,242 134,242 24

Table 14: Within-hour Balancing Reserves Incremental (MW) Period BPA Hydro Non-BPA Hydro Non-BPA Thermal 14 October 900 584 562 November 900 748 711 December 900 782 768 January 900 929 816 February 900 763 702 March 900 797 738 April 1-15 400 719 672 April 16-30 400 719 672 May 400 912 910 June 400 810 799 July 900 15 750 958 August 1-15 900 797 640 August 16-31 900 797 640 September 900 716 662 14 These balancing reserves were not assigned for this analysis. 15 BPA s DEC reserve requirements of 400 megawatts extend through the end of July but the analysis in this report incorrectly assumed that the July reserve requirement was 900 megawatts. It was determined that rerunning all of the studies to include this correction was not warranted. 25

Table 15: Within-hour Balancing Reserves Decremental (MW) Period BPA Hydro Non-BPA Hydro Non-BPA Thermal October 900 662 786 November 900 899 1,264 December 900 687 1,073 January 900 751 908 February 900 728 955 March 900 690 899 April 1-15 900 713 942 April 16-30 900 713 942 May 900 748 1,044 June 900 723 898 July 900 629 811 August 1-15 900 609 872 August 16-31 900 609 872 September 900 746 910 FUTURE ASSESSMENTS The Council will continue to assess the adequacy of the region s power supply. This task is becoming more challenging because planners must now focus on satisfying not only winter energy needs but also summer energy needs and capacity needs year round. Continued development of variable generation resources, combined with changing patterns of electricity demand have added complexity to the task of successfully maintaining an adequate power supply. For example, regional planners have had to reevaluate methods to quantify and plan for balancing reserve needs. In light of these changes, the Council is in the process of enhancing its adequacy model to reflect real life operations and to address capacity issues. Another emerging concern is the lack of access to supplies for some utilities due to insufficient transmission or due to other factors. For the current adequacy assessment, the Northwest 26

region is split into two subsections 16 in which only the major east-to-west transmission lines are modeled. Similarly, only the major Canadian-U.S. and Northwest-to-Southwest interties are modeled. The Council is hoping to address these issues in future adequacy assessments. Also, at some point, uncertainties surrounding the change in Canadian flood control operations in 2024 and the effects of a potentially renegotiated Columbia River Treaty will have to be addressed. But besides these issues, the Council s latest power plan identifies the following action items related to adequacy assessments: RES-8 COUN-3 COUN-4 COUN-5 COUN-6 COUN-8 COUN-11 ANLYS-4 ANLYS-22 ANLYS-23 Adaptive Management Annual Resource Adequacy Assessments Review the regional resource adequacy standard Review the RAAC assumptions regarding availability of imports Review the methodology used to calculate the adequacy reserve margins used in the Regional Portfolio Model Review the methodology used to calculate the associated system capacity contribution values used in the Regional Portfolio Model Participate in and track WECC [adequacy] activities Participate in efforts to update and model climate change data Review and enhancement of peak load forecasting GENESYS Model Redevelopment Enhance the GENESYS model to improve the simulation of hourly hydroelectric system operations Issues identified in 2016 by the Council s Resource Adequacy Advisory Committee to consider for next year s assessment include those listed below: Rec-1 Rec-2 Rec-3 Review methodology of the hybrid load forecast used for the 2021 adequacy assessment, in particular how peak loads are forecast Provide an hourly forecast for energy efficiency savings. Investigate how to incorporate uncertainty in EE savings into the adequacy assessments 16 The dividing line between the east and west areas of the region (for modeling purposes) is roughly the Cascade mountain range. 27

Rec-4 Rec-5 Rec-6 Rec-7 Rec-8 Rec-9 Investigate availability of regional and extra-regional market supplies during periods of stress (supply shortages) Investigate the availability of fuel during periods of stress, especially for resources without firm fuel contracts. Investigate the availability of the interties that connect the NW with regions that may provide market supplies. Consider adding maintenance schedules and forced outages. Explore ways to incorporate the effects of climate change into the adequacy assessments. Should assessments only include the effects of recent temperature years or is there a way to adjust historic temperature profiles to account for climate change? Explore how an energy imbalance market might affect adequacy assessments. Investigate ways to incorporate an EIM into the analysis. Review the use of standby resources in the adequacy assessments, in particular how demand response is modeled. The algorithms in the standby resource post processor should be incorporated into the GENESYS model. DR should be dispatched based on price. How do we deal with existing DR, assuming that its impacts have been captured (somewhat) in the load forecast? Not all of the action items and recommendations listed above will be addressed and resolved before the next adequacy assessment, which is tentatively scheduled for release in May of 2017. However, any enhancements that can be made and tested in time for the next assessment will be implemented. Thus, it continues to be important to isolate the effects of modeling changes on the LOLP from the effects of changes in loads and resources. q:\jf\2021 adequacy\2021 adequacy state of the system report.docx 28

Northwest Regional Forecast of Power Loads and Resources 2017 through 2026 April 2016

Special thanks to System Planning Committee members and utility staff that provided us with this information. Electronic copies of this report are available on the website www..org 101 SW Main Street, Suite 1605 Portland, OR 97204 503.294.1259 2016 Northwest Regional Forecast

Table of Contents Executive Summary... 1 Overview... 9 Planning Area... 9 Northwest Region Requirements and Resources... 10 Annual Energy Table 1... 10 2016-2017 Monthly Energy Table 2... 11 Winter Peak Table 3... 12 Summer Peak Table 4... 13 Northwest New and Existing Resources... 14 Recently Acquired Resources Table 5... 14 Committed New Supply Table 6... 15 Demand Side Management Programs Table 7... 16 Planned Resources Table 8... 17 Planned Resources Table 9... 17 Northwest Utility Generating Resources Table 10... 18 Independent Owned Generating Resources Table 11... 31 Report Procedures... 33 Load Estimates... 33 Demand Side Management... 34 Generating Resources... 35 Thermal and Other Renewable Resources... 36 New and Future Resources... 37 Contracts... 37 Utilities Included in Northwest Regional Forecast Table 12... 38 Definitions... 39 2016 Northwest Regional Forecast

2016 Northwest Regional Forecast

MW/MWa 2016 Northwest Regional Forecast Executive Summary The Northwest Regional Forecast (Forecast) is a compilation of Northwest utilities expected loads and resources through 2026. This annual supply and demand snapshot serves as a barometer for the region s electric power system. Modest load growth expectations, PURPA renewables coming online, and aggressive energy efficiency acquisitions continue to be the theme for the Northwest power sector. The Forecast examines the Northwest utilities power picture at an aggregate level. Individual utilities have different load profiles, risk tolerance and challenges than the region as a whole. Still, looking at the big picture reveals trends in the Northwest energy world. And while winter peak continues to show the largest deficit using the Forecast s planning criteria, summer peak is a growing concern, especially if fewer non-firm resources are available in the summer as compared to winter. Expected load growth remains low Idled smelters keep loads down In 2015 the Northwest s last aluminum giant, Alcoa, announced that it would be idling its regional smelters. The smelters operation is largely hinged on the global price of aluminum. Increased supply in China has pushed the commodity price to low levels in recent years. This lost load has pulled down regional demand expectations for winter peak and annual energy. Summer forecasted loads start in-line with last year s forecast and then grow slightly faster. 1 Load forecasts up and down 34,000 32,000 Winter peak 30,000 28,000 26,000 Summer peak 24,000 2016 report 2015 report 22,000 Annual energy 20,000 2016-17 2018-19 2020-21 Figure 1 1 The forecasted loads reflect expected (1-in-2) weather conditions and savings from projected energy efficiency efforts. 2016 Northwest Regional Forecast 1

Annual energy (MWa) 5y annual energy yearly growth Varying degrees of growth across region On average, regional annual energy load growth is projected at 0.7% per year through 2021. Winter peak load is also forecast at 0.7% while summer peak is 1%. A look at annual energy load growth for individual utilities shows some of them forecasting growth in excess of 1% per year, whereas others are forecasting load decay. Utilities growing faster than 1% are typically a smaller utility expecting a significant new load. Reset on annual energy and winter peak 8% 6% 4% 2% 0% -2% Utility growth varies 6.16% -1.02% Figure 2 Looking at past reports, firm annual energy and winter peak requirement forecasts (load + contracted exports) have continued to start from a lower point than the previous year, implying decreasing need for annual energy and winter peak supply. 23,000 22,000 Annual load requirements reset 21,000 20,000 2012 2013 2014 2015 2016 2012-13 2014-15 2016-17 2018-19 Figure 3 The starting point for the 2016 annual energy requirements forecast is down nearly 1,000 MWa from the 2012 Forecast. This trend is not found in the summer peak forecasts which continue to trend as expected. Resource mix in transition The firm power supply in the Forecast includes hydro at critical water levels, existing utility owned/contracted generating facilities, long-term imports and committed future resources. The Forecast s planning metrics do not include non-firm resources. 2016 Northwest Regional Forecast 2

Monthly energy (MWa) Annual average energy (MWa) Wealth of carbon free resources Largely thanks to the hydropower system, the Northwest has a wealth of CO 2 free power resources. The Forecast assumes critical water conditions for planning purposes, but in any given year the hydro system can generate significantly more power. When the region has more precipitation and generates more hydropower, it relies less on other dispatchable resources, which are largely thermal. This in turn leads to lower CO 2 emissions. The hydro system s generation output under various water conditions, along with other firm carbon free resources stacked on top, are shown below. 25,000 20,000 15,000 10,000 5,000 - Clean hydro supports carbon free mix 2016-17 Firm requirements Critical Average Hydro Nuclear Wind Other CO2 free Figure 4 Hydro and thermal resources work together Although the region s power system provides significant amounts of carbon free power, due to variations in hydro, wind and other CO 2 free resource generation, dispatchable thermal resources are relied upon to fill the gap, even during the highest of water years. Hydro output varies by month The shape of Northwest hydro generation and energy load varies month by month. During higher 24,000 18,000 water years the extra hydro 12,000 generation is largely found in the 6,000 winter, spring and early summer - Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul months. In the late summer and Average hydro Critical hydro Load (2016/17) early fall the difference in Figure 5 generation between critical and average water is less appreciable. This is largely due to the lack of storage on the Northwest s hydro system and the natural snowpack-driven, runoff pattern. 2016 Northwest Regional Forecast 3

MW/MWa Cumulative annual energy (MWa) There are yearly and seasonal variations with wind in the Northwest as well. Wind production tends to be at its highest in the spring/early summer, which combined with hydro, can create a regional energy surplus in those months. Region aggressively acquiring energy efficiency Notable energy efficiency 1,200 2015 forecast 2016 forecast 1,000 800 600 400 200 0 2016-17 2018-19 2020-21 The Forecast s numbers show a region actively pursuing energy efficiency savings as a resource. One reason Northwest load growth has slowed is the thousands of megawatts of energy efficiency savings utilities and others have captured. Utilities expect to achieve additional annual electric energy savings of nearly 1,100 MWa in the next six years, slightly more than last year s Forecast. Once market transformation and codes and standards are accounted for this number will grow. Figure 6 The sun also rises in the Northwest Looking at committed resources, Idaho Power expects nearly 400 MW of nameplate capacity solar within the year via PURPA, and Portland General Electric s natural gas unit Carty is scheduled to be online in 2016. Some hydro system upgrades and PURPA wind in the next few years round out the picture. 1,000 800 600 400 Committed resources stack up In addition, around 2,000 MW of planned resources are identified by utilities to meet future demand. These projects have not been sited or licensed and thus, not included in the Forecast s load/resource tabulations. More details can be found in Tables 8 and 9 Planned Resources of the report. 200 0 Nameplate (MW) Annual energy (MWa) Winter peak (MW) Summer peak (MW) Natural gas Solar Hydro upgrades Wind Figure 7 2016 Northwest Regional Forecast 4

MW Demand response growing to meet peaks Today, the Northwest has hundreds of megawatts of demand response on call. This resource is largely found in the eastern part of the region in the form of irrigation interruption. On the west side, utilities are eyeing this capacity resource as well, with nearly 150 MW of new winter programs scheduled to come on line in the next few years. Resource retirements ahead In the next decade over 2,000 MW of dispatchable capacity, in the form of coal units, are slated to retire. Up first are the planned retirements of Boardman and Centralia Unit 1, scheduled for the end of 2020. Further down the road Centralia Unit 2 is slated to go offline at the end of 2025, and Valmy has been dropped from Idaho Power s preferred portfolio at the end of 2025 (although its retirement is not certain). These retirements occur within the Forecast s horizon. Resource availability for meeting peak capacity and energy needs could be impacted if these dispatchable units are not replaced with resources of similar operating characteristics. 40000 Regional need follows resource retirement 2012 report 2016 report Attention on peak needs 35000 Winter peak is focus 30000 Although winter peak need has been trending down the past five years, it remains the most acute need in this year s Forecast. In 2012 the estimated one-hour peak need for January 2013 was about 3,000 MW. Today that gap is closer to 1,000 MW for January 2017 and grows to over 4,000 in 2021 based the Forecast s planning criteria. 2 This 3,000 MW increase by 2021 is in part due to increased planning margins, 25000 20000 2013, 12% margin expected load growth and the retirement of the Boardman power plant. Figure 8 2017, 12% margin Firm resources Load and exports Load, exports and planning margin 2021, 16% margin It is worth noting that the 2,000 MW decrease in need from 2013 to 2017 is due to a roughly 1,300 MW drop in firm obligations and a 700 MW increase in firm resources. 3 2 1-in-2 load, critical water, utility firm resources and contracts, and 12% planning margin growing 1% a year. 3 Power plant Carty (440 MW) and Port Westward 2 (220 MW) along with a reshuffling of a few contracts. 2016 Northwest Regional Forecast 5

Peak Capacity (MWh) Assumptions can drive seasonal adequacy concerns The assumptions for non-firm resources vary between organizations and can drive which season is of greatest concern. 4 To help shed light on the potential for utilizing non-firm resources, this year s Forecast provides a bookend that shows how the firm power supply can be augmented if generation from independent power producers (IPPs), spot market imports, and additional hydropower (when water supply exceeds critical condition levels), are available. A snapshot of the load/resource picture for winter and summer peak with a potential set of non-firm resources layered on is shown below. Firm resources come from the Forecast, assumptions for available generation from Northwest IPPs and market imports are from the Council s 2015 Resource Adequacy Assessment, and the estimate of additional hydro generation from average water conditions is derived from the 2015 BPA White Book. 5 As noted, the season of greatest concern could be winter or summer depending on non-firm resource assumptions. Market assumptions shape seasonal picture Winter 40,000 Summer 30,000 20,000 2017 2018 2019 2020 2021 2017 2018 2019 2020 2021 Planned thermal resources (nameplate) Market imports (RAAC) Firm resources (NRF) Average hydro (incremental, White Book) Northwest IPPs (RAAC) Firm requirements (NRF with planning margin) Figure 9 4 For example, BPA assumes full IPP availability year round, whereas the Council de-rates IPP s in the summer 5 Firm requirements include contracted exports and a planning margin that starts at 12% and grows 1% per year 2016 Northwest Regional Forecast 6

Future is a little foggy Load may not be business as usual Although this report predicts slow load growth, there are a number of possible new loads that could increase the use of electricity in the Northwest. While some of these possible loads are already baked into the Forecast s figures, specific sectors could see greater than predicted growth. Additionally, the possibility of methanol plants in the Northwest could bring large scale industrial load growth to the region. On the other hand, there are a number of programs that could pull load forecasts down further. These are factored into the report to some extent, but there is a chance they have been underestimated. Public policy changing the power supply landscape Although adequacy has been the driver behind some recent power plant builds in the Northwest, public policy, has played a large role as well. This will likely continue into the future with implementation of existing and new policies, and could change the needs of the power system. State renewable portfolio standards have brought thousands of megawatts of variable energy resources to the Northwest and greater Western Interconnection. This has led to greater concerns regarding system flexibility. In addition, the retirement of Boardman and Centralia power plants, which are due in part to carbon driven public policy, may result in the construction of replacement resources. Beyond existing policies there are additional rules and regulations on the drawing board on both a state and federal level. With each new policy there is a level of uncertainty until the policy is finalized and implements. Going forward will continue to keep an eye on new policy developments and ensure members are aware of how they may impact the power system and need for power. Reading the tea leaves is not the only organization that examines projected need for power. The Bonneville Power Administration and the Northwest Power & Conservation Council also conduct regular Northwest supply and demand studies. At a high level they both peg winter capacity as the area of chief concern. 2016 Northwest Regional Forecast 7

BPA s 2015 White Book Regional Picture The BPA White Book uses various methods to assess regional need for power, including critical water planning similar to the Forecast. One major difference between the White Book and the Forecast is the treatment of power supply from Northwest Independent Power Producers the White Book assumes that 100 percent of PNW regional uncommitted IPP generation is available to serve regional loads. 6 The White Book found the region to be constrained regarding January 120 hour capacity need starting in 2019, even with the inclusion of IPP resources. 7 The Forecast does not have a 120 hour metric to compare. Looking at 1 hour capacity need, with all IPP resources available, the White Book sees a deficit starting in 2021. 8 One key driver of the 2021 deficit is the retirement of Boardman and Centralia Unit 1. Council s Resource Adequacy Advisory Committee 2015 Assessment Each year the Council conducts a regional probabilistic five year out loss-of-load study with the goal of having a less than 5% annual chance of a supply based power outage. The Assessment for year 2020 also featured a six year outlook to examine the region after the coal unit retirements. They found a region that was adequate in year 2020, but inadequate in 2021, with the chief concern being winter capacity. 9 6 Bonneville Power Administration, 2015 White Book Summary Document, Jan 2016, p. 37. IPPs are ~ 3,100 MW 7 Bonneville Power Administration, 2015 White Book Summary Document, p. 42 8 Bonneville Power Administration, 2015 White Book Technical Appendix Volume 2, Capacity Analysis, p. 352 9 Northwest Power and Conservation Council, Pacific Northwest Power Supply Adequacy Assessment and State of the System Report, May 2015, p 11 2016 Northwest Regional Forecast 8

Overview Each year the Northwest Regional Forecast compiles utilities 10-year projections of electric loads and resources which provide information about the region s need to acquire new power supply. The Forecast is a comprehensive look at the capability of existing and new electric generation resources, long-term firm contracts, expected savings from demand side management programs and other components of electric demand for the Northwest. This report presents estimates of annual average energy, seasonal energy and winter and summer peak capability in Tables 1 through 4 of the Northwest Region Requirements and Resources section. These metrics provide a multi-dimensional look at the Northwest s need for power and underscore the growing complexity of the power system. Northwest generating resources are shown by fuel type. Existing resources include those resources listed in Tables 5, 6, 10 and 11. Table 5, Recently Acquired Resources highlights projects and supply that became available most recently. Table 6, Committed New Supply lists those generating projects where construction has started, as well as contractual arrangements that have been made for providing power at a future time. Table 10, Northwest Utility Generating Resources is a comprehensive list of generating resources that make up the electric power supply for the Pacific Northwest that are utility-owned or utility contracted. Table 11, Independent Owned Generating Resources lists generating projects owned by independent power producers and located in the Northwest. In addition, utilities have demand side management programs in place to reduce the need for generating resources. Table 7, Demand Side Management Programs provides a snapshot of utilities expected savings from these programs for the next ten years. Table 8, Planned Resources is a compilation of what utilities have reported in their individual integrated resource plans to meet future need. Planning Area The Northwest Regional Planning Area is the area defined by the Pacific Northwest Electric Power Planning and Conservation Act. It includes: the states of Oregon, Washington and Idaho; Montana west of the Continental Divide; portions of Nevada, Utah, and Wyoming that lie within the Columbia River drainage basin; and any rural electric cooperative customer not in the geographic area described above, but served by BPA on the effective date of the Act. 2016 Northwest Regional Forecast 9